debian-mirror-gitlab/doc/user/clusters/applications.md
2020-06-23 00:09:42 +05:30

56 KiB

stage group info
Configure Configure To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/engineering/ux/technical-writing/#designated-technical-writers

GitLab Managed Apps

GitLab provides GitLab Managed Apps, a one-click install for various applications which can be added directly to your configured cluster.

These applications are needed for Review Apps and deployments when using Auto DevOps.

You can install them after you create a cluster.

Installing applications

Applications managed by GitLab will be installed onto the gitlab-managed-apps namespace.

This namespace:

  • Is different from the namespace used for project deployments.
  • Is created once.
  • Has a non-configurable name.

To see a list of available applications to install. For a:

Install Helm first as it's used to install other applications.

NOTE: Note: As of GitLab 11.6, Helm will be upgraded to the latest version supported by GitLab before installing any of the applications.

The following applications can be installed:

With the exception of Knative, the applications will be installed in a dedicated namespace called gitlab-managed-apps.

NOTE: Note: Some applications are installable only for a project-level cluster. Support for installing these applications in a group-level cluster is planned for future releases. For updates, see the issue tracking progress.

CAUTION: Caution: If you have an existing Kubernetes cluster with Helm already installed, you should be careful as GitLab cannot detect it. In this case, installing Helm via the applications will result in the cluster having it twice, which can lead to confusion during deployments.

Helm

  • Introduced in GitLab 10.2 for project-level clusters.
  • Introduced in GitLab 11.6 for group-level clusters.

Helm is a package manager for Kubernetes and is required to install all the other applications. It is installed in its own pod inside the cluster which can run the helm CLI in a safe environment.

NOTE: Note: Installing Helm as a GitLab-managed App behind a proxy is not supported, but a workaround is available.

cert-manager

Introduced in GitLab 11.6 for project- and group-level clusters.

cert-manager is a native Kubernetes certificate management controller that helps with issuing certificates. Installing cert-manager on your cluster will issue a certificate by Let's Encrypt and ensure that certificates are valid and up-to-date.

The chart used to install this application depends on the version of GitLab used. In:

If you have installed cert-manager prior to GitLab 12.3, Let's Encrypt will block requests from older versions of cert-manager.

To resolve this:

  1. Uninstall cert-manager (consider backing up any additional configuration).
  2. Install cert-manager again.

GitLab Runner

  • Introduced in GitLab 10.6 for project-level clusters.
  • Introduced in GitLab 11.10 for group-level clusters.

GitLab Runner is the open source project that is used to run your jobs and send the results back to GitLab. It is used in conjunction with GitLab CI/CD, the open-source continuous integration service included with GitLab that coordinates the jobs.

If the project is on GitLab.com, shared Runners are available (the first 2000 minutes are free, you can buy more later) and you do not have to deploy one if they are enough for your needs. If a project-specific Runner is desired, or there are no shared Runners, it is easy to deploy one.

Note that the deployed Runner will be set as privileged, which means it will essentially have root access to the underlying machine. This is required to build Docker images, so it is the default. Make sure you read the security implications before deploying one.

NOTE: Note: The runner/gitlab-runner chart is used to install this application, using a preconfigured values.yaml file. Customizing the installation by modifying this file is not supported.

Ingress

  • Introduced in GitLab 10.2 for project-level clusters.
  • Introduced in GitLab 11.6 for group-level clusters.

Ingress provides load balancing, SSL termination, and name-based virtual hosting out of the box. It acts as a web proxy for your applications and is useful if you want to use Auto DevOps or deploy your own web apps.

The Ingress Controller installed is Ingress-NGINX, which is supported by the Kubernetes community.

NOTE: Note: With the following procedure, a load balancer must be installed in your cluster to obtain the endpoint. You can use either Ingress, or Knative's own load balancer (Istio) if using Knative.

In order to publish your web application, you first need to find the endpoint which will be either an IP address or a hostname associated with your load balancer.

To install it, click on the Install button for Ingress. GitLab will attempt to determine the external endpoint and it should be available within a few minutes.

Determining the external endpoint automatically

Introduced in GitLab 10.6.

After you install Ingress, the external endpoint should be available within a few minutes.

TIP: Tip: This endpoint can be used for the Auto DevOps base domain using the KUBE_INGRESS_BASE_DOMAIN environment variable.

If the endpoint doesn't appear and your cluster runs on Google Kubernetes Engine:

  1. Check your Kubernetes cluster on Google Kubernetes Engine to ensure there are no errors on its nodes.
  2. Ensure you have enough Quotas on Google Kubernetes Engine. For more information, see Resource Quotas.
  3. Check Google Cloud's Status to ensure they are not having any disruptions.

Once installed, you may see a ? for "Ingress IP Address" depending on the cloud provider. For EKS specifically, this is because the ELB is created with a DNS name, not an IP address. If GitLab is still unable to determine the endpoint of your Ingress or Knative application, you can determine it manually.

NOTE: Note: The stable/nginx-ingress chart is used to install this application with a values.yaml file.

Determining the external endpoint manually

If the cluster is on GKE, click the Google Kubernetes Engine link in the Advanced settings, or go directly to the Google Kubernetes Engine dashboard and select the proper project and cluster. Then click Connect and execute the gcloud command in a local terminal or using the Cloud Shell.

If the cluster is not on GKE, follow the specific instructions for your Kubernetes provider to configure kubectl with the right credentials. The output of the following examples will show the external endpoint of your cluster. This information can then be used to set up DNS entries and forwarding rules that allow external access to your deployed applications.

If you installed Ingress via the Applications, run the following command:

kubectl get service --namespace=gitlab-managed-apps ingress-nginx-ingress-controller -o jsonpath='{.status.loadBalancer.ingress[0].ip}'

Some Kubernetes clusters return a hostname instead, like Amazon EKS. For these platforms, run:

kubectl get service --namespace=gitlab-managed-apps ingress-nginx-ingress-controller -o jsonpath='{.status.loadBalancer.ingress[0].hostname}'

For Istio/Knative, the command will be different:

kubectl get svc --namespace=istio-system istio-ingressgateway -o jsonpath='{.status.loadBalancer.ingress[0].ip} '

Otherwise, you can list the IP addresses of all load balancers:

kubectl get svc --all-namespaces -o jsonpath='{range.items[?(@.status.loadBalancer.ingress)]}{.status.loadBalancer.ingress[*].ip} '

NOTE: Note: If EKS is used, an Elastic Load Balancer will also be created, which will incur additional AWS costs.

NOTE: Note: You may see a trailing % on some Kubernetes versions, do not include it.

The Ingress is now available at this address and will route incoming requests to the proper service based on the DNS name in the request. To support this, a wildcard DNS CNAME record should be created for the desired domain name. For example, *.myekscluster.com would point to the Ingress hostname obtained earlier.

Using a static IP

By default, an ephemeral external IP address is associated to the cluster's load balancer. If you associate the ephemeral IP with your DNS and the IP changes, your apps will not be able to be reached, and you'd have to change the DNS record again. In order to avoid that, you should change it into a static reserved IP.

Read how to promote an ephemeral external IP address in GKE.

Pointing your DNS at the external endpoint

Once you've set up the external endpoint, you should associate it with a wildcard DNS record such as *.example.com. in order to be able to reach your apps. If your external endpoint is an IP address, use an A record. If your external endpoint is a hostname, use a CNAME record.

Web Application Firewall (ModSecurity)

Introduced in GitLab 12.7.

A Web Application Firewall (WAF) examines traffic being sent or received, and can block malicious traffic before it reaches your application. The benefits of a WAF are:

  • Real-time security monitoring for your application
  • Logging of all your HTTP traffic to the application
  • Access control for your application
  • Highly configurable logging and blocking rules

Out of the box, GitLab provides you with a WAF known as ModSecurity.

ModSecurity is a toolkit for real-time web application monitoring, logging, and access control. With GitLab's offering, the OWASP's Core Rule Set, which provides generic attack detection capabilities, is automatically applied.

This feature:

  • Runs in "Detection-only mode" unless configured otherwise.

  • Is viewable by checking your Ingress controller's modsec log for rule violations. For example:

    kubectl logs -n gitlab-managed-apps $(kubectl get pod -n gitlab-managed-apps -l app=nginx-ingress,component=controller --no-headers=true -o custom-columns=:metadata.name) modsecurity-log -f
    

To enable WAF, switch its respective toggle to the enabled position when installing or updating Ingress application.

If this is your first time using GitLab's WAF, we recommend you follow the quick start guide.

There is a small performance overhead by enabling ModSecurity. If this is considered significant for your application, you can disable ModSecurity's rule engine for your deployed application in any of the following ways:

  1. Setting the deployment variable AUTO_DEVOPS_MODSECURITY_SEC_RULE_ENGINE to Off. This will prevent ModSecurity from processing any requests for the given application or environment.

  2. Switching its respective toggle to the disabled position and applying changes through the Save changes button. This will reinstall Ingress with the recent changes.

Disabling WAF

Logging and blocking modes

To help you tune your WAF rules, you can globally set your WAF to either Logging or Blocking mode:

  • Logging mode - Allows traffic matching the rule to pass, and logs the event.
  • Blocking mode - Prevents traffic matching the rule from passing, and logs the event.

To change your WAF's mode:

  1. Install ModSecurity if you have not already done so.
  2. Navigate to {cloud-gear} Operations > Kubernetes.
  3. In Applications, scroll to Ingress.
  4. Under Global default, select your desired mode.
  5. Click Save changes.
WAF version updates

Enabling, disabling, or changing the logging mode for ModSecurity is only allowed within same version of Ingress due to limitations in Helm which might be overcome in future releases.

ModSecurity UI controls are disabled if the version deployed differs from the one available in GitLab, while actions at the Ingress level, such as uninstalling, can still be performed:

WAF settings disabled

Updating Ingress to the most recent version enables you to take advantage of bug fixes, security fixes, and performance improvements. To update Ingress application, you must first uninstall it, and then re-install it as described in Install ModSecurity.

Viewing Web Application Firewall traffic

Introduced in GitLab Ultimate 12.9.

You can view Web Application Firewall traffic by navigating to your project's Security & Compliance > Threat Monitoring page.

From there, you can see tracked over time:

  • The total amount of traffic to your application.
  • The proportion of traffic that is considered anomalous by the Web Application Firewall's default OWASP ruleset.

If a significant percentage of traffic is anomalous, it should be investigated for potential threats, which can be done by examining the Web Application Firewall logs.

Threat Monitoring

JupyterHub

  • Introduced in GitLab 11.0 for project-level clusters.
  • Introduced in GitLab 12.3 for group and instance-level clusters.

JupyterHub is a multi-user service for managing notebooks across a team. Jupyter Notebooks provide a web-based interactive programming environment used for data analysis, visualization, and machine learning.

Authentication will be enabled only for project members for project-level clusters and group members for group-level clusters with Developer or higher access to the associated project or group.

We use a custom Jupyter image that installs additional useful packages on top of the base Jupyter. You will also see ready-to-use DevOps Runbooks built with Nurtch's Rubix library.

More information on creating executable runbooks can be found in our Runbooks documentation. Note that Ingress must be installed and have an IP address assigned before JupyterHub can be installed.

NOTE: Note: The jupyter/jupyterhub chart is used to install this application with a values.yaml file.

Jupyter Git Integration

  • Introduced in GitLab 12.0 for project-level clusters.
  • Introduced in GitLab 12.3 for group and instance-level clusters.

When installing JupyterHub onto your Kubernetes cluster, JupyterLab's Git extension is automatically provisioned and configured using the authenticated user's:

  • Name.
  • Email.
  • Newly created access token.

JupyterLab's Git extension enables full version control of your notebooks as well as issuance of Git commands within Jupyter. Git commands can be issued via the Git tab on the left panel or via Jupyter's command line prompt.

NOTE: Note: JupyterLab's Git extension stores the user token in the JupyterHub DB in encrypted format and in the single user Jupyter instance as plain text. This is because Git requires storing credentials as plain text. Potentially, if a nefarious user finds a way to read from the file system in the single user Jupyter instance they could retrieve the token.

Jupyter's Git Extension

You can clone repositories from the files tab in Jupyter:

Jupyter clone repository

Knative

  • Introduced in GitLab 11.5 for project-level clusters.
  • Introduced in GitLab 12.3 for group- and instance-level clusters.

Knative provides a platform to create, deploy, and manage serverless workloads from a Kubernetes cluster. It is used in conjunction with, and includes Istio to provide an external IP address for all programs hosted by Knative.

You will be prompted to enter a wildcard domain where your applications will be exposed. Configure your DNS server to use the external IP address for that domain. For any application created and installed, they will be accessible as <program_name>.<kubernetes_namespace>.<domain_name>. This will require your Kubernetes cluster to have RBAC enabled.

NOTE: Note: The knative/knative chart is used to install this application.

Prometheus

  • Introduced in GitLab 10.4 for project-level clusters.
  • Introduced in GitLab 11.11 for group-level clusters.

Prometheus is an open-source monitoring and alerting system useful to supervise your deployed applications.

GitLab is able to monitor applications automatically, using the Prometheus integration. Kubernetes container CPU and memory metrics are automatically collected, and response metrics are retrieved from NGINX Ingress as well.

To enable monitoring, simply install Prometheus into the cluster with the Install button.

NOTE: Note: The stable/prometheus chart is used to install this application with a values.yaml file.

Crossplane

  • Introduced in GitLab 12.5 for project-level clusters.

Crossplane is a multi-cloud control plane useful for managing applications and infrastructure across multiple clouds. It extends the Kubernetes API using:

  • Custom resources.
  • Controllers that watch those custom resources.

Crossplane allows provisioning and lifecycle management of infrastructure components across cloud providers in a uniform manner by abstracting cloud provider-specific configurations.

The Crossplane GitLab-managed application:

  • Installs Crossplane with a provider of choice on a Kubernetes cluster attached to the project repository.
  • Can then be used to provision infrastructure or managed applications such as PostgreSQL (for example, CloudSQL from GCP or RDS from AWS) and other services required by the application via the Auto DevOps pipeline.

For information on configuring Crossplane installed on the cluster, see Crossplane configuration.

NOTE: Note: alpha/crossplane chart v0.4.1 is used to install Crossplane using the values.yaml file.

Elastic Stack

Introduced in GitLab 12.7 for project- and group-level clusters.

Elastic Stack is a complete end-to-end log analysis solution which helps in deep searching, analyzing and visualizing the logs generated from different machines.

GitLab is able to gather logs from pods in your cluster automatically. Filebeat will run as a DaemonSet on each node in your cluster, and it will ship container logs to Elasticsearch for querying. GitLab will then connect to Elasticsearch for logs instead of the Kubernetes API, and you will have access to more advanced querying capabilities.

Log data is automatically deleted after 30 days using Curator.

To enable log shipping:

  1. Ensure your cluster contains at least 3 nodes of instance types larger than f1-micro, g1-small, or n1-standard-1.
  2. Navigate to {cloud-gear} Operations > Kubernetes.
  3. In Kubernetes Cluster, select a cluster.
  4. In the Applications section, find Elastic Stack and click Install.

NOTE: Note: The gitlab/elastic-stack chart is used to install this application with a values.yaml file.

NOTE: Note: The chart deploys 3 identical Elasticsearch pods which can't be colocated, and each require 1 CPU and 2 GB of RAM, making them incompatible with clusters containing fewer than 3 nodes or consisting of f1-micro, g1-small, n1-standard-1, or *-highcpu-2 instance types.

NOTE: Note: The Elastic Stack cluster application is intended as a log aggregation solution and is not related to our Advanced Global Search functionality, which uses a separate Elasticsearch cluster.

Optional: deploy Kibana to perform advanced queries

If you are an advanced user and have direct access to your Kubernetes cluster using kubectl and helm, you can deploy Kibana manually.

The following assumes that helm has been initialized with helm init.

Save the following to kibana.yml:

elasticsearch:
  enabled: false

filebeat:
  enabled: false

kibana:
  enabled: true
  elasticsearchHosts: http://elastic-stack-elasticsearch-master.gitlab-managed-apps.svc.cluster.local:9200

Then install it on your cluster:

helm repo add gitlab https://charts.gitlab.io
helm install --name kibana gitlab/elastic-stack --values kibana.yml

To access Kibana, forward the port to your local machine:

kubectl port-forward svc/kibana-kibana 5601:5601

Then, you can visit Kibana at http://localhost:5601.

Fluentd

Introduced in GitLab 12.10 for project- and group-level clusters.

Fluentd is an open source data collector, which enables you to unify the data collection and consumption to better use and understand your data. Fluentd sends logs in syslog format.

To enable Fluentd:

  1. Navigate to {cloud-gear} Operations > Kubernetes and click Applications. You will be prompted to enter a host, port and protocol where the WAF logs will be sent to via syslog.
  2. Provide the host domain name or URL in SIEM Hostname.
  3. Provide the host port number in SIEM Port.
  4. Select a SIEM Protocol.
  5. Select at least one of the available logs (such as WAF or Cilium).
  6. Click Save changes.

Fluentd input fields

Future apps

Interested in contributing a new GitLab managed app? Visit the development guidelines page to get started.

Install using GitLab CI/CD (alpha)

Introduced in GitLab 12.6.

CAUTION: Warning: This is an alpha feature, and it is subject to change at any time without prior notice.

This alternative method allows users to install GitLab-managed applications using GitLab CI/CD. It also allows customization of the install using Helm values.yaml files.

Supported applications:

Usage

You can find and import all the files referenced below in the example cluster applications project.

To install applications using GitLab CI/CD:

  1. Connect the cluster to a cluster management project.

  2. In that project, add a .gitlab-ci.yml file with the following content:

    include:
      - template: Managed-Cluster-Applications.gitlab-ci.yml
    
  3. Add a .gitlab/managed-apps/config.yaml file to define which applications you would like to install. Define the installed key as true to install the application and false to uninstall the application. For example, to install Ingress:

    ingress:
      installed: true
    
  4. Optionally, define .gitlab/managed-apps/<application>/values.yaml file to customize values for the installed application.

A GitLab CI/CD pipeline will then run on the master branch to install the applications you have configured. In case of pipeline failure, the output of the Helm Tiller binary will be saved as a CI job artifact.

Important notes

Note the following:

  • We recommend using the cluster management project exclusively for managing deployments to a cluster. Do not add your application's source code to such projects.
  • When you set the value for installed key back to false, the application will be unprovisioned from the cluster.
  • If you update .gitlab/managed-apps/<application>/values.yaml with new values, the application will be redeployed.

Install Ingress using GitLab CI/CD

To install Ingress, define the .gitlab/managed-apps/config.yaml file with:

ingress:
  installed: true

Ingress will then be installed into the gitlab-managed-apps namespace of your cluster.

You can customize the installation of Ingress by defining a .gitlab/managed-apps/ingress/values.yaml file in your cluster management project. Refer to the chart for the available configuration options.

Install cert-manager using GitLab CI/CD

cert-manager is installed using GitLab CI/CD by defining configuration in .gitlab/managed-apps/config.yaml.

cert-manager:

  • Is installed into the gitlab-managed-apps namespace of your cluster.
  • Can be installed with or without a default Let's Encrypt ClusterIssuer, which requires an email address to be specified. The email address is used by Let's Encrypt to contact you about expiring certificates and issues related to your account.

The following configuration is required to install cert-manager using GitLab CI/CD:

certManager:
  installed: true
  letsEncryptClusterIssuer:
    installed: true
    email: "user@example.com"

The following installs cert-manager using GitLab CI/CD without the default ClusterIssuer:

certManager:
  installed: true
  letsEncryptClusterIssuer:
    installed: false

You can customize the installation of cert-manager by defining a .gitlab/managed-apps/cert-manager/values.yaml file in your cluster management project. Refer to the chart for the available configuration options.

Install Sentry using GitLab CI/CD

NOTE: Note: The Sentry Helm chart recommends at least 3GB of available RAM for database migrations.

To install Sentry, define the .gitlab/managed-apps/config.yaml file with:

sentry:
  installed: true

Sentry will then be installed into the gitlab-managed-apps namespace of your cluster.

You can customize the installation of Sentry by defining .gitlab/managed-apps/sentry/values.yaml file in your cluster management project. Refer to the chart for the available configuration options.

We recommend you pay close attention to the following configuration options:

  • email. Needed to invite users to your Sentry instance and to send error emails.
  • user. Where you can set the login credentials for the default admin user.
  • postgresql. For a PostgreSQL password that can be used when running future updates.

NOTE: Note: When upgrading it is important to provide the existing PostgreSQL password (given using the postgresql.postgresqlPassword key) or you will receive authentication errors. See the PostgreSQL chart documentation for more information.

Here is an example configuration for Sentry:

# Admin user to create
user:
  # Indicated to create the admin user or not,
  # Default is true as the initial installation.
  create: true
  email: "<your email>"
  password: "<your password>"

email:
  from_address: "<your from email>"
  host: smtp
  port: 25
  use_tls: false
  user: "<your email username>"
  password: "<your email password>"
  enable_replies: false

ingress:
  enabled: true
  hostname: "<sentry.example.com>"

# Needs to be here between runs.
# See https://github.com/helm/charts/tree/master/stable/postgresql#upgrade for more info
postgresql:
  postgresqlPassword: example-postgresql-password

Install PostHog using GitLab CI/CD

PostHog 🦔 is a developer-friendly, open-source product analytics platform.

To install PostHog into the gitlab-managed-apps namespace of your cluster, define the .gitlab/managed-apps/config.yaml file with:

posthog:
  installed: true

You can customize the installation of PostHog by defining .gitlab/managed-apps/posthog/values.yaml in your cluster management project. Refer to the Configuration section of the PostHog chart's README for the available configuration options.

NOTE: Note: You must provide a PostgreSQL password in postgresql.postgresqlPassword or you will receive authentication errors. See the PostgreSQL chart documentation for more information.

Redis pods are restarted between upgrades. To prevent downtime, provide a Redis password using the redis.password key. This prevents a new password from being generated on each restart.

Here is an example configuration for PostHog:

ingress:
  enabled: true
  hostname: "<posthog.example.com>"

# This will be autogenerated if you skip it. Include if you have 2 or more web replicas
posthogSecret: 'long-secret-key-used-to-sign-cookies'

# Needs to be here between runs.
# See https://github.com/helm/charts/tree/master/stable/postgresql#upgrade for more info
postgresql:
  postgresqlPassword: example-postgresql-password

# Recommended to set this to a value to redis prevent downtime between upgrades
redis:
  password: example-redis-password

NOTE: Note: Support for the PostHog managed application is provided by the PostHog team. If you run into issues, please open a support ticket directly.

Install Prometheus using GitLab CI/CD

Introduced in GitLab 12.8.

Prometheus is an open-source monitoring and alerting system for supervising your deployed applications.

To install Prometheus into the gitlab-managed-apps namespace of your cluster, define the .gitlab/managed-apps/config.yaml file with:

prometheus:
  installed: true

You can customize the installation of Prometheus by defining .gitlab/managed-apps/prometheus/values.yaml in your cluster management project. Refer to the Configuration section of the Prometheus chart's README for the available configuration options.

Install GitLab Runner using GitLab CI/CD

GitLab Runner is installed using GitLab CI/CD by defining configuration in .gitlab/managed-apps/config.yaml.

The following configuration is required to install GitLab Runner using GitLab CI/CD:

gitlabRunner:
  installed: true

GitLab Runner is installed into the gitlab-managed-apps namespace of your cluster.

In order for GitLab Runner to function, you must specify the following:

  • gitlabUrl - the GitLab server full URL (for example, https://example.gitlab.com) to register the Runner against.
  • runnerRegistrationToken - The registration token for adding new Runners to GitLab. This must be retrieved from your GitLab instance.

These values can be specified using CI variables:

  • GITLAB_RUNNER_GITLAB_URL will be used for gitlabUrl.
  • GITLAB_RUNNER_REGISTRATION_TOKEN will be used for runnerRegistrationToken

You can customize the installation of GitLab Runner by defining .gitlab/managed-apps/gitlab-runner/values.yaml file in your cluster management project. Refer to the chart for the available configuration options.

Install Cilium using GitLab CI/CD

Introduced in GitLab 12.8.

Cilium is a networking plugin for Kubernetes that you can use to implement support for NetworkPolicy resources. For more information, see Network Policies.

For an overview, see the Container Network Security Demo for GitLab 12.8.

Enable Cilium in the .gitlab/managed-apps/config.yaml file to install it:

# possible values are gke, eks or you can leave it blank
clusterType: gke

cilium:
  installed: true

The clusterType variable enables the recommended Helm variables for a corresponding cluster type. The default value is blank. You can check the recommended variables for each cluster type in the official documentation:

You can customize Cilium's Helm variables by defining the .gitlab/managed-apps/cilium/values.yaml file in your cluster management project. Refer to the Cilium chart for the available configuration options.

CAUTION: Caution: Installation and removal of the Cilium requires a manual restart of all affected pods in all namespaces to ensure that they are managed by the correct networking plugin.

NOTE: Note: Major upgrades might require additional setup steps, please consult the official upgrade guide for more information.

By default, Cilium will drop all disallowed packets upon policy deployment. The audit mode is scheduled for release in Cilium 1.8. In the audit mode, disallowed packets will not be dropped, and audit notifications will be generated instead. GitLab provides alternative Docker images for Cilium with the audit patch included. You can switch to the custom build and enable the audit mode by adding the following to .gitlab/managed-apps/cilium/values.yaml:

global:
  registry: registry.gitlab.com/gitlab-org/defend/cilium
  policyAuditMode: true

agent:
  monitor:
    eventTypes: ["drop", "audit"]

The Cilium monitor log for traffic is logged out by the cilium-monitor sidecar container. You can check these logs with the following command:

kubectl -n gitlab-managed-apps logs cilium-XXXX cilium-monitor

You can disable the monitor log in .gitlab/managed-apps/cilium/values.yaml:

agent:
  monitor:
    enabled: false

The Hubble monitoring daemon is enabled by default and it's set to collect per namespace flow metrics. This metrics are accessible on the Threat Monitoring dashboard. You can disable Hubble by adding the following to .gitlab/managed-apps/config.yaml:

cilium:
  installed: true
  hubble:
    installed: false

You can also adjust Helm values for Hubble via .gitlab/managed-apps/cilium/hubble-values.yaml:

metrics:
  enabled:
    - 'flow:sourceContext=namespace;destinationContext=namespace'

Install Vault using GitLab CI/CD

Introduced in GitLab 12.9.

Hashicorp Vault is a secrets management solution which can be used to safely manage and store passwords, credentials, certificates and more. A Vault installation could be leveraged to provide a single secure data store for credentials used in your applications, GitLab CI/CD jobs, and more. It could also serve as a way of providing SSL/TLS certificates to systems and deployments in your infrastructure. Leveraging Vault as a single source for all these credentials allows greater security by having a single source of access, control, and auditability around all your sensitive credentials and certificates.

To install Vault, enable it in the .gitlab/managed-apps/config.yaml file:

vault:
  installed: true

By default you will get a basic Vault setup with no scalable storage backend. This is enough for simple testing and small-scale deployments, though has limits to how much it can scale, and as it is a single instance deployment, you will experience downtime when upgrading the Vault application.

To optimally use Vault in a production environment, it's ideal to have a good understanding of the internals of Vault and how to configure it. This can be done by reading the the Vault documentation as well as the Vault Helm chart values.yaml file.

At a minimum you will likely set up:

The following is an example values file (.gitlab/managed-apps/vault/values.yaml) that configures Google Key Management Service for auto-unseal, using a Google Cloud Storage backend, enabling the Vault UI, and enabling HA with 3 pod replicas. The storage and seal stanzas below are examples and should be replaced with settings specific to your environment.

# Enable the Vault WebUI
ui:
  enabled: true
server:
  # Disable the built in data storage volume as it's not safe for Hight Availablity mode
  dataStorage:
    enabled: false
  # Enable High Availability Mode
  ha:
    enabled: true
    # Configure Vault to listen on port 8200 for normal traffic and port 8201 for inter-cluster traffic
    config: |
      listener "tcp" {
        tls_disable = 1
        address = "[::]:8200"
        cluster_address = "[::]:8201"
      }
      # Configure Vault to store its data in a GCS Bucket backend
      storage "gcs" {
        path = "gcs://my-vault-storage/vault-bucket"
        ha_enabled = "true"
      }
      # Configure Vault to automatically unseal storage using a GKMS key
      seal "gcpckms" {
         project     = "vault-helm-dev-246514"
         region      = "global"
         key_ring    = "vault-helm-unseal-kr"
         crypto_key  = "vault-helm-unseal-key"
      }      

Once you have successfully installed Vault, you will need to initialize the Vault and obtain the initial root token. You will need access to your Kubernetes cluster that Vault has been deployed into in order to do this. To initialise the Vault, get a shell to one of the Vault pods running inside Kubernetes (typically this is done by using the kubectl command line tool). Once you have a shell into the pod, run the vault operator init command:

kubectl -n gitlab-managed-apps exec -it vault-0 sh
/ $ vault operator init

This should give you your unseal keys and initial root token. Make sure to note these down and keep these safe as you will need them to unseal the Vault throughout its lifecycle.

Install JupyterHub using GitLab CI/CD

Introduced in GitLab 12.8.

JupyterHub is installed using GitLab CI/CD by defining configuration in .gitlab/managed-apps/config.yaml as follows:

jupyterhub:
  installed: true
  gitlabProjectIdWhitelist: []
  gitlabGroupWhitelist: []

In the configuration:

  • gitlabProjectIdWhitelist restricts GitLab authentication to only members of the specified projects.
  • gitlabGroupWhitelist restricts GitLab authentication to only members of the specified groups.
  • Specifying an empty array for both will allow any user on the GitLab instance to sign in.

JupyterHub is installed into the gitlab-managed-apps namespace of your cluster.

For JupyterHub to function, you must set up an OAuth Application. Set:

  • "Redirect URI" to http://<JupyterHub Host>/hub/oauth_callback.
  • "Scope" to api read_repository write_repository.

In addition, the following variables must be specified using CI variables:

CI Variable Description
JUPYTERHUB_PROXY_SECRET_TOKEN Secure string used for signing communications from the hub. Seeproxy.secretToken.
JUPYTERHUB_COOKIE_SECRET Secure string used for signing secure cookies. See hub.cookieSecret.
JUPYTERHUB_HOST Hostname used for the installation. For example, jupyter.gitlab.example.com.
JUPYTERHUB_GITLAB_HOST Hostname of the GitLab instance used for authentication. For example, gitlab.example.com.
JUPYTERHUB_AUTH_CRYPTO_KEY A 32-byte encryption key used to set auth.state.cryptoKey.
JUPYTERHUB_AUTH_GITLAB_CLIENT_ID "Application ID" for the OAuth Application.
JUPYTERHUB_AUTH_GITLAB_CLIENT_SECRET "Secret" for the OAuth Application.

By default, JupyterHub will be installed using a default values file. You can customize the installation of JupyterHub by defining a .gitlab/managed-apps/jupyterhub/values.yaml file in your cluster management project.

Refer to the chart reference for the available configuration options.

Install Elastic Stack using GitLab CI/CD

Introduced in GitLab 12.8.

Elastic Stack is installed using GitLab CI/CD by defining configuration in .gitlab/managed-apps/config.yaml.

The following configuration is required to install Elastic Stack using GitLab CI/CD:

elasticStack:
  installed: true

Elastic Stack is installed into the gitlab-managed-apps namespace of your cluster.

You can check the default values.yaml we set for this chart.

You can customize the installation of Elastic Stack by defining .gitlab/managed-apps/elastic-stack/values.yaml file in your cluster management project. Refer to the chart for the available configuration options.

NOTE: Note: In this alpha implementation of installing Elastic Stack through CI, reading the environment logs through Elasticsearch is unsupported. This is supported if installed via the UI.

Install Crossplane using GitLab CI/CD

Introduced in GitLab 12.9.

Crossplane is installed using GitLab CI/CD by defining configuration in .gitlab/managed-apps/config.yaml.

The following configuration is required to install Crossplane using GitLab CI/CD:

Crossplane:
  installed: true

Crossplane is installed into the gitlab-managed-apps namespace of your cluster.

You can check the default values.yaml we set for this chart.

You can customize the installation of Crossplane by defining .gitlab/managed-apps/crossplane/values.yaml file in your cluster management project. Refer to the chart for the available configuration options. Note that this link points to the documentation for the current development release, which may differ from the version you have installed.

Install Fluentd using GitLab CI/CD

Introduced in GitLab 12.10.

To install Fluentd into the gitlab-managed-apps namespace of your cluster using GitLab CI/CD, define the following configuration in .gitlab/managed-apps/config.yaml:

Fluentd:
  installed: true

You can also review the default values set for this chart in the values.yaml file.

You can customize the installation of Fluentd by defining .gitlab/managed-apps/fluentd/values.yaml file in your cluster management project. Refer to the configuration chart for the current development release of Fluentd for the available configuration options.

NOTE: Note: The configuration chart link points to the current development release, which may differ from the version you have installed. To ensure compatibility, switch to the specific branch or tag you are using.

Install Knative using GitLab CI/CD

To install Knative, define the .gitlab/managed-apps/config.yaml file with:

knative:
  installed: true

You can customize the installation of Knative by defining .gitlab/managed-apps/knative/values.yaml file in your cluster management project. Refer to the chart for the available configuration options.

Here is an example configuration for Knative:

domain: 'my.wildcard.A.record.dns'

If you plan to use GitLab Serverless capabilities, be sure to set an A record wildcard domain on your custom configuration.

Knative Metrics

GitLab provides Invocation Metrics for your functions. To collect these metrics, you must have:

  1. Knative and Prometheus managed applications installed on your cluster.

  2. Manually applied the custom metrics on your cluster by running the following command:

    kubectl apply -f https://gitlab.com/gitlab-org/cluster-integration/cluster-applications/-/raw/02c8231e30ef5b6725e6ba368bc63863ceb3c07d/src/default-data/knative/istio-metrics.yaml
    

Uninstall Knative

To uninstall Knative, you must first manually remove any custom metrics you have added by running the following command:

kubectl delete -f https://gitlab.com/gitlab-org/cluster-integration/cluster-applications/-/raw/02c8231e30ef5b6725e6ba368bc63863ceb3c07d/src/default-data/knative/istio-metrics.yaml

Upgrading applications

Introduced in GitLab 11.8.

The applications below can be upgraded.

Application GitLab version
Runner 11.8+

To upgrade an application:

  1. For a:
  2. Select your cluster.
  3. If an upgrade is available, the Upgrade button is displayed. Click the button to upgrade.

NOTE: Note: Upgrades will reset values back to the values built into the runner chart plus the values set by values.yaml

Uninstalling applications

Introduced in GitLab 11.11.

The applications below can be uninstalled.

Application GitLab version Notes
cert-manager 12.2+ The associated private key will be deleted and cannot be restored. Deployed applications will continue to use HTTPS, but certificates will not be renewed. Before uninstalling, you may wish to back up your configuration or revoke your certificates.
GitLab Runner 12.2+ Any running pipelines will be canceled.
Helm 12.2+ The associated Tiller pod, the gitlab-managed-apps namespace, and all of its resources will be deleted and cannot be restored.
Ingress 12.1+ The associated load balancer and IP will be deleted and cannot be restored. Furthermore, it can only be uninstalled if JupyterHub is not installed.
JupyterHub 12.1+ All data not committed to GitLab will be deleted and cannot be restored.
Knative 12.1+ The associated IP will be deleted and cannot be restored.
Prometheus 11.11+ All data will be deleted and cannot be restored.
Crossplane 12.5+ All data will be deleted and cannot be restored.
Elastic Stack 12.7+ All data will be deleted and cannot be restored.
Sentry 12.6+ The PostgreSQL persistent volume will remain and should be manually removed for complete uninstall.

To uninstall an application:

  1. For a:
  2. Select your cluster.
  3. Click the Uninstall button for the application.

Support for uninstalling all applications is planned for progressive rollout. To follow progress, see the relevant epic.

Troubleshooting applications

Applications can fail with the following error:

Error: remote error: tls: bad certificate

To avoid installation errors:

  • Before starting the installation of applications, make sure that time is synchronized between your GitLab server and your Kubernetes cluster.

  • Ensure certificates are not out of sync. When installing applications, GitLab expects a new cluster with no previous installation of Helm.

    You can confirm that the certificates match via kubectl:

    kubectl get configmaps/values-content-configuration-ingress -n gitlab-managed-apps -o \
    "jsonpath={.data['cert\.pem']}" | base64 -d > a.pem
    kubectl get secrets/tiller-secret -n gitlab-managed-apps -o "jsonpath={.data['ca\.crt']}" | base64 -d > b.pem
    diff a.pem b.pem
    

Error installing managed apps on EKS cluster

If you're using a managed cluster on AWS EKS, and you are not able to install some of the managed apps, consider checking the logs.

You can check the logs by running following commands:

kubectl get pods --all-namespaces
kubectl get services --all-namespaces

If you are getting the Failed to assign an IP address to container error, it's probably due to the instance type you've specified in the AWS configuration. The number and size of nodes might not have enough IP addresses to run or install those pods.

For reference, all the AWS instance IP limits are found in this AWS repository on GitHub (search for InstanceENIsAvailable).